Prediction of bond strength of lightweight concretes by using artificial neural networks

Abstract

In the scope of this study, concrete samples planned to be used as load-bearing concrete were produced by using pumice aggregate and silica fume. Cement was replaced by silica fume, as the mineral additive, by 5 and 10% of its weight. First of all, fresh concrete properties of the produced samples were evaluated. Then, compressive strength tests were conducted on the 28th and 90th days. In addition, pull-out tests were carried out on cubic samples of 150 mm3 on the 90th day so as to detect the reinforcing steel-concrete bond strength. The data obtained at the end of the tests were used as input to the Artificial Neural Networks (ANN) method to predict bond strength values. Bond strength values predicted via the ANN method were found to be close to the bond strength values obtained via tests. In conclusion, it can be quite beneficial to predict the bond strength of normal and lightweight concrete via the ANN method by using a high number of parameters as input. Thus, it will be possible to detect the reinforcing steel-concrete bond strength in a faster and reliable manner and by doing less laboratory work. Key words: Artificial neural networks, structural lightweight concrete, pumice, bond strength.